Print This PostThe Great White Whale in Alzheimer’s disease research is a drug that prevents cognitively unimpaired people with elevated levels of amyloid from developing frank cognitive impairment or dementia. In 2025, there were 72 clinical trials that had at least some participants in this preclinical phase. Preceding these ongoing trials were a number of disappointments. The most well-known of these was the A4 trial of solanezumab in preclinical Alzheimer’s, which targeted amyloid with the aim of slowing cognitive decline. Results showed that the medication was no more efficacious than a placebo over a period of 240 weeks. Other trials targeting inflammation and oxidative stress, like the ADAPT and GEM studies, have been similarly discouraging. This lack of progress raises an inevitable question: Why are such diverse trials in preclinical Alzheimer’s not producing positive results? In this post, we argue that “cognitively normal” trial samples have unmeasured phenotypic heterogeneity that has masked the ability to detect treatment effects.
Before unpacking this argument and highlighting some of our work addressing unmeasured phenotypic heterogeneity in preclinical AD, it is important for us to acknowledge that there are many factors that could have contributed to lackluster preclinical AD trial results. First, it is possible that existing drugs don’t have the right biological target. Second, it may be that drugs are not being delivered at the correct window of biological or clinical disease progression to have an effect; that is, they are either delivered too late or too early. Third, it could be that the time window of trials is simply too short to identify treatment effects in a preclinical phase characterized by slow decline over many years. Fourth, it’s possible that outcome measures in preclinical Alzheimer’s trials are not sensitive enough to detect the subtle declines that emerge in this stage. Finally, it may be that there is unmeasured biological heterogeneity in preclinical Alzheimer’s samples, such as comorbid neuropathologies, that hinders the effectiveness of drugs targeting Alzheimer’s biology.
As we await the results of trials with different biological targets, improved outcome measures, longer time windows, and better measurement of comorbid neuropathologies, there is another barrier to trial success that we can address with existing techniques right now. Specifically, we can better account for phenotypic heterogeneity in preclinical Alzheimer’s that has historically masked our ability to detect treatment effects. In past trials, anyone who did not have a diagnosable clinical condition of mild cognitive impairment or dementia was lumped into a catchall category of “cognitively normal”, “cognitively unimpaired”, or “preclinical Alzheimer’s”. Over the past decade, there has been increasing recognition that the preclinical phase of Alzheimer’s disease is actually highly heterogeneous both in the presence and type of subtle symptoms that emerge.
One study using data from 285 amyloid positive, cognitively unimpaired older adults from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) found that just over half of participants (56%) had no detectable symptoms. Of the remaining sample, 10% had subjective cognitive symptoms, 14% had subtle patterns of low scores on objective cognitive tests, 6% had neurobehavioral symptoms, like depression, anxiety, or apathy, and 13% had some combination of the three symptom types.
This analysis demonstrated that pooling all amyloid positive, cognitively unimpaired individuals into one catchall preclinical group is an oversimplified way to study such a phenotypically diverse set of individuals.
Failing to account for this preclinical phenotypic diversity in Alzheimer’s trials has hindered the ability to detect treatment effects. This point was further illustrated by research conducted in a sample of 747 participants across the spectrum of pre-dementia stages. The authors reported that 46% of cognitively unimpaired participants with objectively defined subtle cognitive difficulties progressed to mild cognitive impairment by four years of follow up, compared with only 17% of cognitively unimpaired individuals with normal neuropsychological test performance. The group with objective subtle cognitive decline also demonstrated more rapid accumulation of amyloid and more pronounced degeneration of the entorhinal cortex and the hippocampus over those 4 years.
These results suggest strongly that an intervention targeting the preclinical phase of Alzheimer’s would be more likely to detect a treatment effect if only individuals with subtle symptoms were selected for participation. An anti-amyloid drug, for example, would not accomplish much in the cognitively unimpaired, completely asymptomatic individuals, who accumulate little amyloid because the drug has limited available target. And at a clinical level, there is much less risk of decline in this group for the drug to protect against. Excluding participants who are completely asymptomatic, whilst selecting participants who are subtly symptomatic could yield an immense biological and clinical advantage for Alzheimer’s trials.
Putting this plan into action would involve clearly operationalizing objective subtle cognitive decline, defined conceptually as a transitional state of reduced cognitive performance that follows “normal cognition” and precedes mild cognitive impairment. A recent review suggested that there are at least six broad approaches for empirically measuring objective subtle cognitive decline, including methods based on patterns of low scores or change in scores across repeated neuropsychological assessments. It remains to be seen which method is preferred, and we are currently in the process of conducting a Delphi consensus study on this very topic. However, one promising method may involve deployment of digital cognitive assessments.
Research has shown that scores from high frequency smartphone-based cognitive tests are correlated with amyloid pathology. Moreover, the absence of practice effects across these repeated mobile cognitive tests appears to be a particularly sensitive marker of preclinical cognitive decline and amyloid deposition. Such tests can be administered to large numbers of potential participants at relatively low expense, enabling rapid selection of cognitively unimpaired individuals with objective subtle cognitive decline for clinical research.
Complementary to this approach are methods using item-based and process-scores. These techniques break down the test into components to reveal underlying neurocognitive mechanisms, such as forgetting, serial memory, and lexical ability. Using these approaches to uncover mechanistic problems can increase sensitivity to Alzheimer’s pathology without necessarily requiring new tests or test re-design. In some cases, these mechanisms may also be universal and apply across languages and cultures. And importantly, they are suitable to low tech medicine solutions available in areas that face health and digital inequalities.
Starting today, we can harness digital technologies and novel psychometric approaches to identify the individuals with preclinical Alzheimer’s who are most at risk for biological and clinical disease progression—those with more advanced subtle symptoms. If successful, we’ll maximize the ability to detect treatment effects, and we may just catch that Great White Whale, landing a drug that can slow or even halt the progression of Alzheimer’s disease before dementia sets in.
Dr Andrew Kiselica
Author
Dr Andrew M. Kiselica is an Associate Professor at the University of Georgia and a board certified clinical neuropsychologist. His research focuses on detecting early cognitive changes in preclinical Alzheimer’s disease, improving assessment methods, and expanding screening in underserved populations. He works closely with the Working Group on Objective Subtle Cognitive Decline in Alzheimer’s Disease, contributing to efforts to better define and measure early cognitive changes that could improve clinical trials.
Ajantha Abey
Beccy Owen
Dr Sam Moxon
Rahul Sidhu